1

Civil War and the Onset of Militarized Interstate Disputes: Comparing

Allied and Non-Allied Countries

How do civil wars impact militarized inter-state dispute initiations (MIDs) and does this

relationship vary between allied and non-allied states? This article offers a new

perspective on the link between domestic and international conflict and provides the first

theoretical and empirical analysis of the differences in the relationship between civil wars

and MID initiations for allied and non-allied states. In line with previous studies, I find

that civil war in either the initiator or the target state makes the onset of MIDs more

likely for both allied and non-allied dyads, regardless of the alliance type. In addition,

non-allied dyads initiate MIDs at a statistically significantly higher rate than allied dyads,

with or without civil wars. However, I also find surprising and perhaps counterintuitive

result that the relationship between the civil war in initiator and target state flips for a

specific type of alliances: alliances that are highly institutionalized and alliances

involving specific mediation/arbitration measures. The coefficient for civil war in the

state that is the target of MID is statistically significantly different and stronger for allied

than for non-allied dyads. In other words, when there is a civil war in the state that is a

target of MID, common alliance ties make the dispute initiation more, not less likely. This

is surprising because we would expect the specific provisions contained in these alliances

to contribute towards making conflict less likely. This finding potentially supports the

theoretical argument advanced by Ronald Krebs, who argues that membership in

multilateral and institutionalized alliances may intensify conflict between smaller

members because the major power member of the alliance guarantees their external

security, thus allowing them to focus on their preexisting disputes and creating a perverse

incentive for conflict.1

In this article, I first discuss the existing literature linking civil and international

conflict. I then put forward the theoretical utility of distinguishing between allied and

non-allied dyads, focusing on some of the ways in which common alliance ties may

modify the link between civil and international conflict. I conclude by providing tentative

1 Ronald R. Krebs, “Perverse Institutionalism: NATO and the Greco-Turkish Conflict,”

2

empirical results between civil war and the initiation of MIDs, dividing the sample into

allied and non-allied dyads. To the best of my knowledge, this is the first work that

explicitly addresses how common alliance ties may impact the relationship between civil

war and MID initiation.

Existing Literature Linking Domestic and International Conflict

Scholars have discussed many possible causal mechanisms that link domestic political

instability in the form of civil war and the initiation of militarized interstate disputes. The

existing research in international relations (IR) has predominantly focused on two causal

pathways linking civil and international conflict: opportunistic attacks and diversionary

wars. Other scholars have recently offered a novel discussion of how specific issues over

which civil war is fought may provide incentives for external actors to intervene, thus

internationalizing a domestic conflict.

Diversionary War Arguments

At its core, diversionary war arguments posit that leaders initiate the use of force abroad

to deflect attention from domestic issues.2 According to these arguments, the loss of

domestic societal support as a result of economic or other problems incentivizes leaders

to use force externally to unite the country against a common foe and prolong their

tenure. The underling logic is that initiating conflict with some “out-group” may increase

cohesion within the “in-group” and that leaders exploit this to bolster domestic support.

As the level of external threat increases, the public downplays the importance of domestic

problems and becomes more likely to support the regime in office. Diversionary war

theories attribute dispute or conflict initiation to the country experiencing domestic

instability. While intra-state conflict may be one domestic issue, diversionary war

arguments also focus on other domestic problems that may result in the loss of support

for the regime in power (e.g., strikes, riots, low approval rating, etc.). As a result, I

review different variants of the diversionary war theory.

Qualitative, state-specific studies have arguably found the strongest evidence for

diversionary war theory. For example, James and Oneal found that U.S. presidents did 2 Jack Levy. “The Diversionary Theory of War: A Critique.” In Handbook of War Studies, edited by Manus I. Midlarsky (New York: Unwin Hyman, 1989).

3

tend to engage in diversionary wars. 3 Morgan and Bickers reconceptualize the

diversionary war theory and argue that only certain domestic problems, “those associated

with a loss of support for the political leadership from within the groups constituting the

ruling coalition,” will lead to diversionary behavior.4 They analyze the behavior of U.S.

presidents from 1953 to 1976 and find that they have been more prone to behave

aggressively in foreign policy when faced with a loss of partisan support, thus finding

empirical support for their argument. Focusing on the Argentine junta’s decision to

launch the invasion of the Falkland Islands, Amy Oakes finds mixed evidence for

diversionary theory of war and argues that the combination of low extractive capacity and

high domestic unrest compelled Argentine leaders to initiate the invasion.5

Scholars have also analyzed regime-specific arguments in light of diversionary

war theory. Analyzing the initiation of force by the challenging states in 180 crises,

Christopher Gelpi argues that leaders have three strategies when faced with domestic

unrest: (1) grant the demands of dissatisfied groups; (2) repress the dissatisfied groups by

force; (3) divert the public’s attention by using force externally.6 He finds empirical

support that democratic leaders respond to domestic unrest by using force internationally,

while authoritarian leaders repress domestic unrest.7 Leeds and Davis’ cross-national

study examines the theories linking domestic political vulnerability to international

disputes, focusing on the relationship between economic decline, the electoral cycle, and

measures of aggressive international action for 18 advanced industrialized democracies

from 1952 to 1988. They find no consistent support for a relationship between

constraining domestic political conditions and international behavior partially because

other countries place fewer international demands on politically vulnerable

3 Patrick James and John R. Oneal, “The Influence of Domestic and International Politics on the President’s Use of Force,” Journal of Conflict Resolution 35, issue 2 (June 1, 1991): 307-332. 4 T. Clifton Morgan and Kenneth N. Bickers, “Domestic Discontent and the External Use of Force,” The Journal of Conflict Resolution 36, no.1 (March, 1992): 34. 5 Amy Oakes, “Diversionary War and Argentina’s Invasion of the Falkland Islands,” Security Studies 15, no.3 (July-September 2006): 431-463. 6 Christopher Gelpi, “Democratic Diversions: Governmental Structure and the Externalization of Domestic Conflict,” The Journal of Conflict Resolution 41, no.2 (April, 1997): 256. 7 Ibid: 255-282.

4

leaders.8 Building on the research of Leeds and Davis, Ross A. Miller uses three

indicators of domestic unrest (changing levels of economic growth, domestic protest, and

rebellion) and also finds little to no empirical support for their relationship to dispute

escalation, for both democratic and autocratic regimes.9 Focusing more closely on the

organizational incentives faced by different institutional actors, Dassel and Reinhardt

have argued that domestic strife will lead to external aggression only when it threatens

the organizational interests of the military.10

As this brief discussion of illustrates, evidence for diversionary war theory is

much stronger in single-case, qualitative studies, as opposed to large, cross-national

quantitative studies.

Opportunistic War Arguments

Contrary to diversionary war theory, the opportunistic war arguments posit that the state

contending with domestic unrest becomes the victim of attack by an outside,

opportunistic state. Already weakened as a result of domestic conflict, these states

become tempting targets of external, opportunistic challengers. More certain of victory,

external actors may provide direct military and economic assistance to the rebel groups

fighting the government or join the conflict directly on the side of the rebels in order to

advance their interests.

External actors may be motivated by a number of instrumental factors, such as

desire for economic gain, territorial irredentism, or the advancement of other

military/strategic interests. In a qualitative study of revolutions, Stephen Walt finds that

revolutionary states become involved in wars with their neighbors because revolutions

often modify the threat calculation for both the revolutionary state and other countries.11,

Stephen Saideman applies Walt’s balance-of-threat theory and offers an additional way in

which states can balance threats: “supporting efforts, particularly those of secessionist

8 Brett Ashley Leeds and David R. Davis, “Domestic Political Vulnerability and International Disputes,” The Journal of Conflict Resolution 41, no.6 (December 1997): 814-834. 9 Ross A. Miller, “Regime Type, Strategic Interaction, and the Diversionary Use of Force,” The Journal of Conflict Resolution 43, no.3 (June 1999): 388-402. 10 Kurt Dassel and Eric Reinhardt, “Domestic Strife and the Initiation of Violence at Home and Abroad,” American Journal of Political Science 43, no.1 (January, 1999): 56-85. 11 Stephen Walt, Revolution and War (Ithaca, NY: Cornell University Press, 1997).

5

movements, to weaken one’s adversary by promoting its dissolution.”12 As a result of

supporting secessionist movements, states improve their security by assisting efforts that

ultimately reduce the aggregate power of their adversaries through prolonging the

conflict.

In other cases, domestic conflict pitting the government against rebels can invite

external participation if outside state has feelings of solidarity with embattled ethnic kin

and desires to come to their assistance. This is another causal logic that focuses on

opportunities for intervention resulting from ethnic solidarity between the intervener and

the rebels fighting the government.13

Notwithstanding these arguments, in a direct rebuttal of the

victimization/opportunism thesis, Peter Trumbore finds that “external actors are much

more likely to use higher levels of force against states that aren't struggling with ethnic

rebellion than ones that are, directly contradicting the victimization thesis so prevalent in

the literature on the international dimensions of ethnic conflict.”14 As a way of explaining

this seemingly counterintuitive finding, Trumbore argues that states dealing with

insurgency are already weakened and that external actors need only issue threats or

provide assistance to rebels in order to achieve their goals.

Civil War-Specific Arguments

Neither diversionary nor opportunistic arguments focus on the specific issues over

which civil war is fought. Rather, these arguments simply posit that domestic political

and economic disturbances, civil war being one of them, provide incentives for states to

either initiate the use of force (i.e., diversion) or to be victims of force (i.e., opportunism).

Other scholars, however, have more specifically examined the relationship between

issues surrounding civil war and inter-state conflicts and militarized disputes.

For example, Gleditsch et al. demonstrate that civil wars increase the risk of

militarized conflict between states and that this increase in the risk of interstate conflict

12 Stephen M. Saideman. The Ties That Divide: Ethnic Politics, Foreign Policy, and International Conflict. (New York: Columbia University Press, 2001): 18. 13 For a great discussion of this argument, see David A. Lake and Donald Rothchild, The International Spread of Ethnic Conflict: Fear, Diffusion, and Escalation (Princeton: Princeton University Press, 1998). 14 Peter F. Trumbore, “Victims or Aggressors? Ethno-Political Rebellion and Use of Force in Militarized Interstate Disputes,” International Studies Quarterly 47, issue 2 (June 2003): 197.

6

“is driven by states’ efforts to affect the outcome of the civil war and, to a lesser extent,

unintended spillovers from those efforts.”15 They find little evidence that the increased

risk of interstate conflict results from opportunistic attacks or diversion. Linking civil to

international conflict through specific mechanisms, Boaz Atzili identifies two ways in

which civil conflict could lead to international conflict.16 The first involves refugee flows

and cross-border insurgency, while the second involves countries intervening to protect

their ethnic brethren across the border. Focusing on the spread of refugees, Salehyan and

Gleditsch find that refugee flows and population movements more broadly are an

additional mechanism by which conflict spreads across regions.17 The specific causal

mechanisms they discuss are the spread of rebel social networks, the facilitation of the

transnational spread of arms, combatants, and ideologies, and the altering of the ethnic

composition of the state.

Alliances, Civil War, and Interstate Militarized Disputes

The existing scholarship has identified a number of causal pathways by which domestic

conflict or instability can provoke inter-state militarized disputes. There is considerable

empirical support for arguments that link civil war or rebellion to international conflict.

As a theoretical matter, and to clarify further discussion, we can identify six distinct

causal logics by which civil war in a given state may become internationalized and lead

to militarized inter-state disputes involving external states:

(1) State engaged in civil conflict initiates MID against external state because the

latter provides assistance to rebels, either direct or indirect. Direct support involves the

provision of economic and military assistance, while indirect support ranges from

allowing the arms to flow through its territory or allowing rebels to cross the border and

use its territory as sanctuary. For example, Salehyan has argued that national boundaries

increase the cost of counterinsurgency for the government because they allow

15 Kristian Skrede Gleditsch, Idean Salehyan, and Kenneth Schultz, “Fighting at Home, Fighting Abroad: How Civil Wars Lead to International Disputes,” Journal of Conflict Resolution 52, no.4 (2008): 479-506. 16 Boaz Atzili, “When Good Fences Make Bad Neighbors: Fixed Borders, State Weakness, and International Conflict,” International Security 31, no.3 (2007): 139-173. 17 Idean Salehyan and Kristian Skrede Gleditsch, “Refugees and the Spread of Civil War,” International Organization 60, issue 2 (April, 2006): 335-366.

7

transnational rebels to use sanctuaries in other states, therefore making it more difficult to

defeat them.18 While the government may pursue rebels across state boundaries (e.g.,

Cambodian attack on Khmer Rouge positions in Thailand or Nicaraguan Sandinista’s

regime attack on the Contras in Honduras), these actions raise the cost of conflict because

the host government reciprocates the attack. Sometimes, the host government may be too

weak to prevent rebels from using its territory as sanctuary, while in other cases it is

actively encouraging and assisting them (e.g., Honduras and Contras in the Nicaraguan

civil war). (2) State contending with civil conflict becomes target of MID by external

state because the latter intervenes directly to support rebels, either because it has common

ethnic/religious ties or because it offers rebels support in the hope that future rebel-

constituted government will be friendlier and/or offer specific concessions.

(3) State contending with civil conflict is significantly weakened and external

state intervenes to satisfy some long-standing claim, irrespective of rebels fighting the

government. This is essentially the opportunism war argument. (4) State contending with

civil conflict initiates MID against external state because it wants to galvanize support for

the regime (i.e., diversionary war argument). (5) States engage in MIDs because of

externalization and spillovers from civil conflict, rather than any deliberate government

policy (e.g., cross-border pursuits of rebels, refugee flows, etc.) In these cases, borders

are usually extremely porous and governments are unable to contain conflict spillover.

(6) Civil war in a given state causes two external states to engage in direct fighting

because they support opposite sides in the conflict. This was the case during the North

Yemeni civil war from 1962 until 1970, during which Egypt and Saudi Arabia engaged in

limited direct fighting because they supported opposing sides.

While the causal logics linking civil to inter-state conflict are theoretically

persuasive and logically consistent, we still lack an understanding if this nexus between

civil war and inter-state conflict varies for allied compared to non-allied states. Almost all

quantitative studies of conflict use alliance as a control variable and find that common

alliance membership generally reduces the likelihood of militarized dispute initiation.19

18 Idean Salehyan, “Transnational Rebels: Neighboring States as Sanctuary for Rebel Groups,” World Politics 59, Number 2 (January 2007): 217-242. 19 For example, see Stuart A. Bremer, “Dangerous Dyads: Conditions Affecting the Likelihood of Interstate War, 1816-1965,” The Journal of Conflict Resolution 36, no.2 (June, 1992): 309-341;

8

In addition, scholars have found that specific alliance provisions, such as commitment to

resolve disputes peacefully or information provision about military capabilities,

contribute to a decrease in militarized dispute initiation between allied states. 20

Disaggregating data into different historical periods, Leeds and Mattes find that alliances

formed in the 1885-1944 period bear no relationship to dispute involvement between

allied members, while alliances formed in the 1945-1991 period have a lower likelihood

of intra-allied conflict.21 Others, however, argue that membership in a common alliance

can sometimes intensify conflict among member states, by changing incentives faced by

small and less powerful members of the alliance.22 Once the great powers in the alliance

provide for the security of small states, these minor powers are free to pursue preexisting

disputes against other alliance members

All of this, however, tells us very little about the similarities and differences in the

conflict patterns between allied and non-allied states when faced with violent domestic

political context. The existence of wars and militarized disputes between countries that

have formal commitments may at first seem counterintuitive and empirically puzzling.

After all, alliances are generally formed either to address members’ external and internal

threats or to manage relations between allied states, whatever they may be. As most

alliances are public documents, they also represent costly signal of states’ intentions that

is observed by other members of the international system. This makes the occurrence of

serious conflict even more puzzling: why would a state sign a formal commitment with

another state and then proceed to use force against it, despite the potential reputational

and other costs? Put differently, what motivates a state to use force against its ally?

Bruce Russett, John R. Oneal, and David R. Davis, “The Third Leg of the Kantian Tripod for Peace: International Organizations and Militarized Disputes, 1950-1985,” International Organization 52, Issue 3 (Summer 1998): 441-467; Paul Huth, Standing Your Ground: Territorial Disputes and International Conflict (Ann Arbor: University of Michigan Press, 1998) 20 See Andrew G. Long, Timothy Nordstrom, and Kyeonghi Baek, “Allying for Peace: Treaty Obligations and Conflict Between Allies,” The Journal of Politics 69, no.4 (November 2007): 1103-1117; et al.; David H. Bearce, Kristen M. Flanagan, and Katherine M. Floros, “Alliances, Internal Information, and Military Conflict Among Member-States,” International Organization 60, no.3 (Summer 2006): 595-625. 21 Brett Ashley Leeds and Michaela Mattes, “Alliance Politics During the Cold War: Aberration, New World Order, or the Continuation of History?” Conflict Management and Peace Science 24 (2007): 183-199. 22 Ronald Krebs, “Perverse Institutionalism: NATO and the Greco-Turkish Conflict.”

9

The first step in answering this question is to establish a basic but nonetheless

useful point. The fighting between allies, just like fighting between non-allied states,

results from some underlying dispute or disagreement that exists between them. A mix of

common and conflicting interests characterizes relations between all states, even those in

the most cohesive alliances. At any point in their relationship, countries in general and

allies in particular may have a plethora of issues over which they disagree. Put

differently, alliance generally does not imply harmonious relationship between states.

The disagreements that exist between allies may sometimes result in fighting, joint

alliance membership notwithstanding. Of course, alliances do not cover all the issues

over which states could disagree and new issues may arise after the alliance is formed

that can cause disagreement and conflict where there wasn’t one before. For example, at

the time NATO was formed, the United States and Turkey could not foresee that they

would come dangerously close to fighting each other over a disagreement about the

political solution to the Syrian conflict.

A closer examination of the dyads’ that form alliances reveals that they usually

contain elements that are most highly correlated with conflict. For example, majority of

alliances are formed between states that are geographically contiguous or in which at

least one alliance member is a major power.23 Research in IR has clearly shown that

geographic contiguity and major power status, among other things, are factors most

highly correlated with the existence of militarized disputes and wars.24 In addition, major

wars and militarized disputes with high casualties are a relatively rare occurrence. Most

countries, including allies, are at peace most of the time.

While it may be true that, on average, common alliance membership and specific

alliance provisions reduce intra-allied conflict, what happens when members are faced

with domestic rebellion that profoundly alters the incentives faced by relevant actors? In

the event of internal conflict, external states, either allied or non-allied, could

theoretically choose from seven distinct strategic options: (1) Intervene directly to

23 I will have specific numbers on this. 24 For more on this, see Senese, Paul D. and John A. Vasquez. “A Unified Explanation of Territorial Conflict: Testing the Impact of Sampling Bias, 1919-1992.” International Studies Quarterly, vol.47/issue 2 (2003): 275-298.; Vasquez, John. The War Puzzle Revisited. (Cambridge University Press: 2009.); Huth, Paul K. Standing Your Ground: Territorial Disputes and International Conflict. (Ann Arbor: University of Michigan Press, 1996).

10

support the government; (2) Intervene directly to support rebels; (3) Provide explicit

economic or military assistance to the government, without overt intervention; (4)

Provide explicit economic or military assistance to rebels, without overt intervention; (5)

Provide indirect assistance to the government (e.g., sanctuary, transit for weapons, etc.);

(6) Provide indirect assistance to rebels; (7) Do nothing. Of course, the external state

could also opt for a combination of any of these policies. The intervention itself,

therefore, can be either to support the incumbent government, oppose it by providing

assistance to rebels, or remain neutral. These policies also carry different costs for the

intervening government. As Gleditsch has argued, “direct intervention in conflicts in

other states, especially on the side of the rebels, constitutes a serious violation of that

state’s sovereignty and often entails significant costs to the intervening state.”25

How do common alliance ties affect the strategic choices that external states face?

Theoretically, membership in a common alliance could make it more likely that the state

intervenes directly on behalf of the government because it wants to preserve the alliance

and defeat rebels, who may be interested in realigning the government with other states.

Indeed, domestic revolutions often change states’ foreign policy in fundamental ways and

increase overall uncertainty in the political system. For example, the Soviet Communist

revolution dramatically reoriented Moscow’s foreign policy, leading the Soviet Union to

abandon World War I efforts and opt for a separate armistice with Germany. Iran’s

revolution in 1979 was a pivotal event in the Middle Eastern politics, leading Tehran to

abandon all existing alliances and reorient its foreign policy away from the United States.

The underlying theoretical point is that internal political unrests are more likely to get

allies involved by the simple fact that allies are more invested in each other’s political

situation.

Allies could also intervene directly against the incumbent government with which

they share alliance ties if they are unhappy with the regime and rebel victory promises

stronger alliance ties in the future. In this case, the external state is unhappy with its ally’s

government and exploits domestic unrest to change the regime and install a more

favorable group. For example, following the Sandinista revolution in Nicaragua in 1979,

25 Kristian Skrede Gleditsch, “Transnational Dimensions of Civil War,” Journal of Peace Research 44, no.3 (2007): 296.

11

the United States and Honduras- Nicaragua’s formal Rio Pact allies- have actively

supported the internal opposition to the regime (i.e., the Contras), with the latter even

engaged in limited direct fighting. In addition, amidst a Costa Rican civil war in 1948,

Costa Rica and Nicaragua- formal allies- engaged in a series of militarized disputes

because the former accused the latter of allowing Costa Rican exiles to operate from

Nicaraguan territory with impunity. To further illustrate this, it is useful to make a

theoretical distinction between inter-state alliances and inter-regime alliances. When two

countries sign an alliance, country A may commit to defend the territorial integrity of

country B. At the same time, however, it does not mean that country A is committed to

defending the regime of country B. As a result, the relationship between two regimes may

deteriorate and, while they still remain formally allied, one country could actively work

to undermine its ally’s regime. Indeed, this was the case with many U.S. interventions in

Central America throughout the Cold War. In addition, the contemporary alliance

between China and North Korea is a good empirical illustration of this. While China

would most likely defend the North Korean state against any external attack, whether it

would intervene to support Kim Jong Un’s regime against domestic insurrection is an

entirely different matter. Indeed, it is not implausible that in case of a civil war in North

Korea, Beijing may intervene overtly to support the opposition, in the hope that new

leadership may be more pro-China.26 Thus, while the inter-state alliance remains solid,

inter-regime alliance may be more fragile.

This begs an important question: Why does common alliance membership matter?

The existence of a formal agreement makes fighting between allies different from

fighting between non-allied states. All militarized disputes and wars between states carry

certain risks and opportunities for participants, regardless of whether they are allied or

not. Defeat may mean loss of territory, resources, political independence, or even

elimination from the state system. Victory may entail acquisition of more territory,

resources, or at the extreme global domination. Notwithstanding these similarities,

conflict between states with formal agreements carries implications that are absent in

non-allied fighting.

26 I acknowledge that the likelihood of this happening is extremely low. This example is only meant to illustrate the broader theoretical point.

12

For example, the obvious difference is that allies risk rupturing the relationship

and destroying the alliance if conflict or disagreements escalate to a certain level,

something not present in the relationship between non-allied states. The possibility of

destroying the alliance or, worse, leading the partner to defect to a hostile state is a cost

exclusive to the conflict between allied states. This is especially so if alliance members

have alternative options for alignment or if they are strong enough to provide for their

own security. A dispute that escalates to war and large-scale fighting that involves

significant casualties will certainly break the alliance, as evidenced by the war between

Austria and Prussia in 1866 or Italy and Albania in 1939. Therefore, in addition to the

costs and risks of fighting, states fighting their allies also have to include the very real

possibility of losing the alliance partner. Assuming that the state values the ally for some

reason (why would it sign the alliance otherwise?), the costs of war are higher than they

otherwise would be because of the risk of losing the ally. This risk may be even bigger in

large, multilateral alliances in which the fighting may be bilateral, but it risks alienating

other alliance partners. For example, Albania decided to withdraw from the Warsaw Pact

and realign itself more closely with China following the Soviet invasion of

Czechoslovakia in 1968. The loss of Albania was an additional cost that Soviets bore in

addition to the material cost of fighting against another ally, Czechoslovakia.

In addition, fighting one’s ally carries reputational costs that are absent in fighting

between non-allied states. When other states observe conflict between alliance members,

they may infer that those states are unreliable partners, thus diminishing their utility as

allies. Indeed, if state A uses force against a state to whom it is formally allied, how can it

credibly commit peaceful relations to other states or other alliance partners? This logic is

nicely summarized by Michaela Mattes and Greg Vonnahme, “Attacking another state or

aiding an aggressor are already likely to be viewed negatively by the international

community but doing so in disregard of an existing formal international commitment

increases the chance that sanctions are imposed.”27 Not all of the implications, however,

are necessarily negative. For example, by acting aggressively against one’s ally, a

country signals its willingness to accept higher costs, which in turn can improve its

27 Michaela Mattes and Greg Vonnahme, “Contracting for Peace: Do Nonaggression Pacts Reduce Conflict?” The Journal of Politics 72, no.4 (2010): 930.

13

perception of resolve among other states. The reputational implication of fighting is the

second important difference in the fighting between allied and non-allied states.

I follow Leeds et al. and define alliances as “written agreements, signed by

official representatives of at least two independent states, that include promises to aid a

partner in the event of military conflict, to remain neutral in the event of conflict, to

refrain from military conflict with one another, or to consult/cooperate in the event of

international crises that create a potential for military conflict.”28 This definition includes

any formal agreement between at least two states and encompasses defense pacts, offense

pacts, neutrality pacts, and consultation pacts. As Leeds and Savun argue, there are

conceptual problems with including non-aggression pacts in the definition of alliances,

“because pure non-aggression pacts require no active coordination, their formation and

termination are governed by different processes.”29 In addition, non-aggression pacts also

include countries that are likely to fight each other (why would they otherwise sign a

non-aggression pact?).30 As a result of this, I exclude exclusively non-aggression pacts

from the analysis.

Research Design

For all of the tests that I conduct, the unit of analysis is directed dyad-year, with each

dyad appearing twice in the sample. I’ve restricted the sample to politically relevant dyad

years, to include territorially contiguous states, allied states, or dyads in which at least

one member is a major power, as defined by the Correlates of War (COW) data set. The

analysis spans the 1816-2010 period and the binary dependent variable is the onset of

MID in which at least one side uses force (i.e., MIDs of level 4 or 5). All of the models

are pooled logit models and robustness checks have been performed using year fixed

effects. As this study analyzes MID onsets, the dependent variable is 1 only for cases in

which state A initiates a new MID against state B, or vice versa. The key independent

variable is the occurrence of civil war in either the initiator or the target. The alliance data

28 Leeds, Brett Ashley et al. (2002). 29 Leeds and Savun (2007): 125. 30 For more on non-aggression pacts, see Michaela Mattes and Greg Vonnahme, “Contracting for Peace: Do Non-Aggression Pacts Reduce Conflict,” The Journal of Politics, vol.72/no.4 (2010): 925-938; Yonatan Lupu, Paul Poast, “Team of Former Rivals: A Multilateral Theory of Non-Aggression Pacts,” Journal of Peace Research, vol.53/issue 3 (2016): 344-358.

14

came from Alliance Treaty Obligations and Provisions (ATOP data set).31 The MID data

and the civil war data came from the Correlates of War (COW) data set.32 I also ran a

robustness check using Uppsala civil war data from 1946-2010 and the results are

broadly consistent.

Because the alliance data is yearly, I studied every case to make sure that the MID

happened after the alliance was formed and after civil war onset. The latter was done for

non-allied dyads as well to remove cases in which MID happened before the official

onset of civil war. Fortunately, there is data on the specific date of the MID as well as the

data on the specific temporal range for the alliance. Within alliances, I compare

institutionalized alliances to ones that are not. It is hypothesized that heavily

institutionalized alliances may be more prone to managing conflicts between allied dyads,

especially if it emanates from internal instability.

Descriptive Statistics

To fully understand the conditions under which formally allied countries engage

in MIDs, I collected narratives of every MID between formally allied states.33 There are

324 unique instances in which formally allied countries initiated MIDs that reached the

level of use of force, from 1816-2010. Out of this total, there are 120 instances (37%) of

MID force initiations during civil war in either the initiator or the target. Out of 120, 18

are instances in which both the initiator and the target were involved in a civil war at the

time of the MID initiations. For non-allied dyads, there are 598 instances of MID force

initiations during civil war in either the initiator or the target Out of this 598, 82 are

instances in which both the initiator and the target were involved in a civil war.

Not all MIDs between states, however, are nefarious in nature. For example,

following a coup d’état in 1964, France intervened in Gabon to restore the legitimate

government, in accordance with Article III of the defense pact between the two countries. 31 Leeds, Brett Ashley, Jeffrey M. Ritter, Sara McLaughlin Mitchell, and Andrew G. Long. 2002. Alliance Treaty Obligations and Provisions, 1815-1944. International Interactions 28: 237-260. 32 Zeev Maoz, Paul L. Johnson, Jasper Kaplan, Fiona Ogunkoya, and Aaron Shreve 2018. The Dyadic Militarized Interstate Disputes (MIDs) Dataset Version 3.0: Logic, Characteristics, and Comparisons to Alternative Datasets, Journal of Conflict Resolution. 33 To ensure consistency in narratives, I relied on a single source that describes each MID. The source is Douglas M. Gibler. International Conflicts, 1816-2010: Militarized Interstate Dispute Narratives, Volume I and II (Rowman & Littlefield, MD: 2018).

15

In addition, the U.S. threatened intervention in Haiti in 1994 to restore the government

led by Aristide, who was overthrown in a coup d’état in 1993. COW data set classifies

these as inter-state disputes when, in fact, they were interventions to support the allied

regime and did not threaten countries’ territorial integrity. In other cases, MIDs were

more nefarious in nature, such as Honduran and U.S. attempts to overthrow the

democratically elected and formally allied Sandinista government in Nicaragua. While

Honduras and the U.S. never had territorial ambitions vis-à-vis Nicaragua, their efforts

were aimed at undermining the regime.

After carefully studying each of the cases, I divide them into three categories: pro-

allied intervention, anti-allied intervention, and neutral fighting. To qualify as a pro-ally

intervention, there has to be evidence that the MID resulted from a direct military

intervention on behalf of the government that is fighting internal opposition or that the

intervention was conducted to support/restore the previous government that was violently

overthrown in a coup d’état or other forms of internal political instability. To qualify as

an anti-ally intervention, there has to be evidence that the MID resulted from direct

financial or military support for rebels fighting the existing legitimate allied government

or from a direct conflict between allies over policy/territorial disagreements or because

one ally has territorial or other aspirations against its ally. Fighting can range from

limited MIDs over policy disagreements to all-out attacks and territorial annexations. I

code as neutral all of the fishing/border disputes and other cases in which fighting is

contained and relatively minor.

Table 1 classifies each of the 324 cases across two dimensions: 1.) Alliance ties,

ranging from a formal defense pact to a neutrality/non-aggression pact; 2.)

Pro/anti/neutral intervention. Each cell in the table shows the number of cases that fit in

each of the categories. For example, there were 16 MIDs in which one ally intervened to

support the other allied regime (e.g., France in Gabon, the U.S. in Haiti). As we can see,

most of the inter-allied MIDs consisted of “neutral” fighting, which is classified as

border/fishing disputes and other relatively minor and contained fighting. The large

number of minor MIDs somewhat undermines the large-N studies of the relationship

between war and other variables that rely on the MIDs data set.34 Indeed, one of the goals

34 This is through no fault of the MID data set itself.

16

of this paper is to further unpack this relationship and get a better understanding of the

kind of fighting states are engaged in. Perhaps most interesting is the 53 anti-allied

interventions in defensive and offensive alliances. This is a relatively big number and it

includes cases in which one ally directly attacked the territorial integrity or political

independence of its fellow allied state.

Alliance ties Defensive/Offensive

Alliance

Neutrality/Non-

Aggression Pacts

Consultation Pacts

Pro-Allied

Intervention

16

Anti-Allied

Intervention

53 6 3

Neutral Fighting 191 29 7

Empirical Analysis

The existing scholarship has already established a strong and positive relationship

between civil war occurrence and MID initiation.35 However, no study has compared

whether the effect varies for allied and non-allied states. Table 2 presents a simple cross-

tabulation, comparing the frequency of MID initiations for allied and non-allied states.

Table 2

MID Initiations between Allied and Non-allied States

In Politically Relevant Dyad Years, 1816-2010.

_______________________________________________________________________

Allied States % Non Allied States %

_______________________________________________________________________

MID Onset 324 0.4 1,569 0.8

No Mid Onset 91,086 99.6 184,676 99.2

Chi-Squared 165.92***

35 See especially Gleditsch et al (2008).

17

________________________________________________________________________

*** Significant at 1 percent.

As Table 2 shows, while overall MID initiation rate is low for both allied and non-allied

states, non-allied states initiate MIDs at twice the rate of allied states (0.8% compared to

0.4%). The difference is statistically significant at the 0.01 level. Table 2 compares MID

initiations between allied and non-allied states, limiting the sample to cases in which

either side A (i.e., the initiator) or side B (i.e., the target) experience civil war at the time

of the MID onset. The results are similar to Table 3 and non-allied states initiate MIDs at

almost twice the rate of allied states and the difference is statistically significant at the

0.01 level.

Table 3

Civil Wars and MID Initiations between Allied and Non-allied States

In Politically Relevant Dyad Years, 1816-2010.

_______________________________________________________________________

Allied States % Non Allied States %

_______________________________________________________________________

MID Onset 120 0.8 598 1.6

No Mid Onset 14,454 99.2 36,435 98.4

Chi-Squared 37.81***

________________________________________________________________________

*** Significant at 1 percent.

The results from the two tables demonstrate that common alliance ties reduce the

frequency of MID onsets for allied states, in both the full sample and the sample limited

to civil war occurrences. However, does the relationship change when we conduct

multivariate analysis and control for a number of factors associated with conflict?

In addition to the two independent variables (i.e., civil war in initiator and target),

I include a number of control variables. Territorial contiguity is a dummy variable,

indicating whether the two states were sharing a land border or being separated by no

more than 400 miles of water. Capabilities ratio is the log of the ratio of the stronger

18

state’s capabilities over the weaker state’s. Major power is a dummy variable and it has a

value of 1 if either member of the dyad was considered a major power. Joint Democracy

is 1 if the dyad is jointly democratic. Cold War is 1 if the relevant years are 1948-1989,

generally considered the beginning and the end of the Cold War. Foreign Policy

Similarity is indicated by the S-score in three different computational measures: global

weighted score using all states (S_weighted_global), score between initiator and system

leader using countries in the relevant region (S_leader1), and the score between target

and system leader using countries in the relevant region (S_leader2). To control for time

dependence, I include the number of years since the last MID in the dyad, entered in

linear form and as a cubic spline.36 I also plan to add variables that predict civil war and

one variable to indicate the degree of dyadic economic interdependence. I lag the values

of the civil war variable by one year, for both initiator and target states.

The first model (Table 4) is a logit model comparing the MID onset between non-

allied and allied dyads. To ease the interpretation of the coefficients, I report the marginal

change in the probability of a MID onset for each variable, holding all other variables at

their means. As is clear, the marginal change in civil war in initiator state is positive and

statistically significant for both non-allied and allied dyads. Civil war in target state is

only significant for non-allied dyads. The baseline predicted probability of MID onset for

non-allied dyads is 0.36%, holding all variables at their means. The same value for allied

dyads is 0.07%. For non-allied dyads, the effect of civil war in initiator or target raises

the predicted probability of MID onset by about 83% over the baseline predicted

probability. For allied dyads, the presence of civil war in initiator increases the predicted

probability of MID onset by about 128% over the baseline predicted probability, a

substantively significant effect.

One of the goals of this article is to compare whether the key independent

variables of interest operate differently among allied and non-allied states. For this, I

conducted a Wald test for the equivalence of civil war coefficients in initiator and target.

In both cases, I fail to reject the null hypothesis that the coefficients are different at

conventional significance levels (χ 2=1.98, p=0.16 for civil war in initiator; χ 2=0.34,

p=0.56 for civil war in target).

36 Beck, Katz, and Tucker 1998. I do not report these results.

19

Table 4

Logit Model of MID Onset, 1816-2010.

(1) (2)

VARIABLES Non-allied dyads Allied dyads

Civil War in Initiator 0.003*** 0.0009***

(0.0005) (0.0003)

Civil War in Target 0.003*** 0.0004

(0.0006) (0.0002)

Ln (Capability Ratio) -0.0009*** -0.0001***

(0.00008) (0.00004)

Major Power 0.002*** 0.0006***

(0.0003) (0.0002)

Joint Democracy -0.002*** -0.0008***

(0.0005) (0.0002)

Contiguity -0.001*** -0.0005***

(0.00009) (0.043)

Cold War 0.002*** 0.635***

(0.0004) (0.00007)

S- global score -0.001*** -0.0005**

(0.0004) (0.0002)

S- initiator score -0.001 0.0011***

(0.0007) (0.0004)

S- target score 0.002** -0.0007

(0.196) (0.0004)

Predicted probability at mean 0.0036*** 0.0007***

(0.0002) (0.0001)

N 93,021 45,690

Note: Robust standard errors in parentheses. Time dependence variables were included in

20

the estimation of the models but are not reported in the table.

*** p

21

Table 5

Logit Model of MID Onset for Non-Allied Dyads and

Institutionalized Allied Dyads, 1816-2001.

(1) (2)

VARIABLES Non-allied dyads Institutionalized allied dyads

Civil War in Initiator 0.600*** 0.396

(0.091) (0.232)

Civil War in Target 0.496*** 1.21***

(0.096) (0.207)

Ln (Capability Ratio) -0.290*** -0.207***

(0.024) (0.063)

Major Power 0.334*** 0.367

(0.109) (0.246)

Joint Democracy -0.442*** -0.548**

(0.150) (0.270)

Contiguity -0.321*** -0.750***

(0.019) (0.048)

Cold War 0.429*** 0.261***

(0.095) (0.172)

S- global score -0.484*** -0.290

(0.116) (0.290)

S- initiator score -0.451** 1.56***

(0.195) (0.545)

S- target score 0.580** -0.666

(0.195) (0.613)

Constant -0.693***

(0.171)

-1.37***

(0.374)

22

N 93,021 54,533

Note: Robust standard errors in parentheses. Time dependence variables were included in

the estimation of the models but are not reported in the table.

*** p

23

Table 6

Logit Model of MID Onset for Non-Allied Dyads and

Allied Dyads with arbitration/mediation measures, 1816-2001.

(1) (2)

VARIABLES Non-allied dyads Allied dyads with

arbitration/mediation

Civil War in Initiator 0.600*** 0.497**

(0.091) (0.233)

Civil War in Target 0.496*** 1.36***

(0.096) (0.217)

Ln (Capability Ratio) -0.290*** -0.183***

(0.024) (0.062)

Major Power 0.334*** 0.583**

(0.109) (0.271)

Joint Democracy -0.442*** -0.782**

(0.150) (0.324)

Contiguity -0.321*** -0.705***

(0.019) (0.049)

Cold War 0.429*** 0.506***

(0.095) (0.192)

S- global score -0.484*** -0.313

(0.116) (0.304)

S- initiator score -0.451** 1.48**

(0.195) (0.545)

S- target score 0.580** -0.931

(0.195) (0.724)

Constant -0.693*** -1.54***

24

(0.171) (0.409)

N 93,021 48,082

Note: Robust standard errors in parentheses. Time dependence variables were included in

the estimation of the models but are not reported in the table.

*** p

25

REFERENCES Atzili, Boaz. “When Good Fences Make Bad Neighbors: Fixed Borders, State Weakness, and International Conflict.” International Security 31, no.3 (2007): 139-173. Bearce, David H., Kristen M. Flanagan, and Katherine M. Floros. “Alliances, Internal Information, and Military Conflict Among Member-States.” International Organization 60, no.3 (Summer 2006): 595-625. Bremer , Stuart A.“Dangerous Dyads: Conditions Affecting the Likelihood of Interstate War, 1816-1965.” The Journal of Conflict Resolution 36, no.2 (June, 1992): 309-341; Dassel, Kurt and Eric Reinhardt. “Domestic Strife and the Initiation of Violence at Home and Abroad.” American Journal of Political Science 43, no.1 (January, 1999): 56-85. Gelpi, Christopher. “Democratic Diversions: Governmental Structure and the Externalization of Domestic Conflict.” The Journal of Conflict Resolution 41, no.2 (April, 1997): 255-282. Gleditsch, Kristian Skrede. “Transnational Dimensions of Civil War.” Journal of Peace Research 44, no.3 (2007): 293-309. Gleditsch, Kristian Skrede, Idean Salehyan, and Kenneth Schultz. “Fighting at Home, Fighting Abroad: How Civil Wars Lead to International Disputes.” Journal of Conflict Resolution 52, no.4 (2008): 479-506. Huth, Paul. Standing Your Ground: Territorial Disputes and International Conflict (Ann Arbor: University of Michigan Press, 1998) James, Patrick and John R. Oneal. “The Influence of Domestic and International Politics on the President’s Use of Force.” Journal of Conflict Resolution 35, issue 2 (June 1, 1991): 307-332. Krebs, Ronald R. “Perverse Institutionalism: NATO and the Greco-Turkish Conflict.” International Organization 53, no. 2 (1999): 343-377. Lake, David A. and Donald Rothchild. The International Spread of Ethnic Conflict: Fear, Diffusion, and Escalation (Princeton: Princeton University Press, 1998). Leeds, Brett Ashley and David R. Davis, “Domestic Political Vulnerability and International Disputes,” The Journal of Conflict Resolution 41, no.6 (December 1997): 814-834. Leeds, Brett Ashley and Michaela Mattes. “Alliance Politics During the Cold War: Aberration, New World Order, or the Continuation of History?” Conflict Management and Peace Science 24 (2007): 183-199. Levy, Jack. “The Diversionary Theory of War: A Critique.” In Handbook of War Studies, edited by Manus I. Midlarsky (New York: Unwin Hyman, 1989). Long, Andrew G., Timothy Nordstrom, and Kyeonghi Baek. “Allying for Peace: Treaty Obligations and Conflict Between Allies.” The Journal of Politics 69, no.4 (November 2007): 1103-1117.

26

Lupu, Yonatan and Paul Poast. “Team of Former Rivals: A Multilateral Theory of Non-Aggression Pacts.” Journal of Peace Research, vol.53/issue 3 (2016): 344-358. Mattes, Michaela and Greg Vonnahme. “Contracting for Peace: Do Nonaggression Pacts Reduce Conflict?” The Journal of Politics 72, no.4 (2010): 925-938. Miller, Ross A. “Regime Type, Strategic Interaction, and the Diversionary Use of Force.” The Journal of Conflict Resolution 43, no.3 (June 1999): 388-402. Morgan, T. Clifton and Kenneth N. Bickers. “Domestic Discontent and the External Use of Force.” The Journal of Conflict Resolution 36, no.1 (March, 1992): 25-52. Oakes, Amy. “Diversionary War and Argentina’s Invasion of the Falkland Islands.” Security Studies 15, no.3 (July-September 2006): 431-463. Russet, Bruce, John R. Oneal, and David R. Davis. “The Third Leg of the Kantian Tripod for Peace: International Organizations and Militarized Disputes, 1950-1985.” International Organization 52, Issue 3 (Summer 1998): 441-467. Saideman, Stephen M. The Ties That Divide: Ethnic Politics, Foreign Policy, and International Conflict. (New York: Columbia University Press, 2001). Salehyan, Idean and Kristian Skrede Gleditsch. “Refugees and the Spread of Civil War.” International Organization 60, issue 2 (April, 2006): 335-366. Salehyan, Idean. “Transnational Rebels: Neighboring States as Sanctuary for Rebel Groups.” World Politics 59, Number 2 (January 2007): 217-242. Senese, Paul D. and John A. Vasquez. “A Unified Explanation of Territorial Conflict: Testing the Impact of Sampling Bias, 1919-1992.” International Studies Quarterly, vol.47/issue 2 (2003): 275-298. Trumbore, Peter F. “Victims or Aggressors? Ethno-Political Rebellion and Use of Force in Militarized Interstate Disputes.” International Studies Quarterly 47, issue 2 (June 2003): 183-201. Vasquez, John. The War Puzzle Revisited. (Cambridge University Press: 2009.) Walt, Stephen. Revolution and War (Ithaca, NY: Cornell University Press, 1997).